Employing Machine Learning Models to Predict Return Loss Precisely in 5G Antenna

Abstract

To meet 5G requirements, designing an optimal antenna is challenging due to numerous design factors. Conventional electromagnetic modeling simulators require excessive time and processing power during the antenna design process. Machine learning (ML), an innovative technology, can be used in the domain of antenna design with favorable performance and can resolve problems that the previous conventional methods cannot. The main goal of this work is to create an antenna that operates at 28 GHz, which is a significant 5G band for the 5G futuristic infrastructure revolution, and to predict the return loss of an antenna using some machine learning models like K-Nearest Neighbor (KNN), Extreme Gradient Boosting (XG-Boost), Decision Tree (DT), and Random Forest (RF). On comparing results, all models perform well with over 83% accuracy. However, the Random Forest model predicts return loss with higher accuracy at 90% and lower MSE and MAE values of 1.99 and 0.827, respectively. Moreover, this antenna holds potential for 5G applications and can be efficiently optimized using a machine learning approach, saving valuable time.